Face recognition processing method and device, computer device and storage medium

Abstract

The application relates to a face recognition processing method and device, computer equipment and a storage medium. The method comprises the following steps: taking the collection angle of the first face information of a target user as a first angle, and extracting the first angle face feature of the target user from the first face information; performing similarity evaluation on the first angle face feature and a plurality of reference face features in a face feature library to obtain a first angle score of the target user; when the first angle score is lower than a score threshold, collecting second face information of the target user based on a second angle different from the first angle to obtain a second angle face feature; performing weighted fusion on the first angle score and a second angle score obtained based on the second angle face feature to obtain a comprehensive score of the target user; and when the comprehensive score is not lower than the score threshold, determining that the target user meets a face recognition condition and performing face recognition processing on the target user. The method can improve the face recognition success rate.

Description

Technical Field

[0001] This application relates to the field of facial recognition technology, and in particular to a facial recognition processing method, apparatus, computer device, and storage medium. Background Technology

[0002] With the development of biometric technology, facial recognition technology, which features non-contact collection, non-contact identification, and long-distance recognition, has been widely used for identity authentication.

[0003] In existing technologies, cameras or webcams are typically used to capture images or video streams containing human faces, and facial feature information is extracted from the images or video streams to perform facial recognition. Based on the results of facial recognition, identity information is confirmed, thereby performing identity authentication.

[0004] However, in existing facial recognition methods, if the user's posture deviates, resulting in a significant difference between the collected facial information and the pre-stored facial information, the success rate of facial recognition will be low. Summary of the Invention

[0005] Therefore, it is necessary to provide a facial recognition processing method, device, computer equipment, computer-readable storage medium, and computer program product that can improve the success rate of facial recognition in order to address the above-mentioned technical problems.

[0006] Firstly, this application provides a facial recognition processing method. The method includes:

[0007] Acquire first facial information of the target user, take the acquisition angle of the first facial information as the first angle, and extract the first angle facial features of the target user from the first facial information.

[0008] The similarity between the first-angle facial features and multiple benchmark facial features in the facial feature database is evaluated to obtain the first-angle score of the target user.

[0009] When the score from the first angle is lower than the score threshold, the second facial features of the target user are obtained based on the second facial information collected from the second angle. The second angle is different from the first angle.

[0010] The first angle score is weighted and fused with the second angle score obtained based on the second angle facial features to obtain the comprehensive score of the target user. When the comprehensive score is not lower than the score threshold, the target user is determined to meet the facial recognition conditions, and facial recognition processing is performed on the target user.

[0011] In one embodiment, prior to acquiring the first facial information collected from the target user, the process includes:

[0012] Identify multiple identifiable users who meet the facial recognition criteria;

[0013] For each identifiable user, acquire the facial features of the identifiable user from multiple acquisition angles;

[0014] For each acquisition angle, the facial features corresponding to the acquisition angle are used as the reference facial features for the acquisition angle, thus obtaining the reference facial features of the user at each of the multiple acquisition angles.

[0015] A facial feature library is constructed based on multiple baseline facial features for each identifiable user.

[0016] In one embodiment, each extracted facial feature includes multiple feature points; the first angle facial feature is compared with multiple benchmark facial features in the facial feature library to obtain the target user's first angle score, which includes:

[0017] All benchmark facial features in the facial feature database that were sampled at the same angle as the first angle were used as benchmark features to be compared.

[0018] For each benchmark feature to be compared, multiple feature points in the first angle facial feature are compared with multiple feature points in the benchmark feature to be compared, and the feature point similarity score between the first angle facial feature and the benchmark feature to be compared is obtained.

[0019] The highest feature point similarity score among multiple feature point similarity scores is used as the target user's first angle score.

[0020] In one embodiment, when the overall score is not lower than a score threshold, the target user is determined to meet the facial recognition conditions. After performing facial recognition processing on the target user, the process includes:

[0021] The first-angle facial features and the second-angle facial features of the target user are respectively used as the new facial features of the target user;

[0022] For each new facial feature, in the target user's facial feature library, the reference facial feature with the same acquisition angle corresponding to the new facial feature is used as the reference facial feature to be supplemented for the new facial feature.

[0023] From the multiple feature points of the newly added facial features, select the difference feature points that differ from the feature points of the corresponding baseline facial features to be supplemented;

[0024] Based on the acquisition angle corresponding to the newly added facial features, the difference feature points corresponding to the newly added facial features are used as auxiliary facial features for the corresponding acquisition angle, and the auxiliary facial features are added to the target user's facial feature library.

[0025] In one embodiment, the target user's facial feature library stores auxiliary facial features and baseline facial features of the target user; each auxiliary facial feature includes at least one differential feature point; adding the auxiliary facial features to the target user's facial feature library includes:

[0026] For each acquisition angle, the total amount of auxiliary facial features corresponding to the acquisition angle of the target user is obtained from the target user's facial feature database.

[0027] The acquisition angles where the total data volume exceeds the data volume threshold are taken as acquisition angles to be adjusted, and the difference feature points in multiple auxiliary facial features corresponding to the acquisition angles to be adjusted are compared.

[0028] For each auxiliary facial feature corresponding to the acquisition angle to be adjusted, the comparison result is determined as the total number of different feature points;

[0029] Auxiliary facial features that do not meet the total number requirement are deleted, and the baseline facial features of the target user at the acquisition angle to be adjusted are updated based on the auxiliary facial features with the fewest total number.

[0030] In one embodiment, when the overall score is not lower than a score threshold, the target user is determined to meet the facial recognition conditions. After performing facial recognition processing on the target user, the method further includes:

[0031] Dynamic behavior detection of target users is performed within a preset effective time period;

[0032] When a target user meets the dynamic behavior detection conditions within the valid time, the target user is authenticated based on the identity information obtained by facial recognition processing.

[0033] Secondly, this application also provides a facial recognition processing device. The device includes:

[0034] The facial feature extraction module is used to acquire first facial information collected from the target user, take the acquisition angle of the first facial information as the first angle, and extract the first angle facial features of the target user from the first facial information.

[0035] The scoring module is used to evaluate the similarity between the first-angle facial features and multiple benchmark facial features in the facial feature library to obtain the first-angle score of the target user.

[0036] The second angle acquisition module is used to obtain the second angle facial features of the target user based on the second facial information collected from the target user from the second angle when the first angle score is lower than the score threshold. The second angle is different from the first angle.

[0037] The comprehensive score acquisition module is used to perform weighted fusion of the first angle score and the second angle score obtained based on the second angle facial features to obtain the comprehensive score of the target user. When the comprehensive score is not lower than the score threshold, it is determined that the target user meets the facial recognition conditions, and facial recognition processing is performed on the target user.

[0038] Thirdly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to perform the following steps:

[0039] Acquire first facial information of the target user, take the acquisition angle of the first facial information as the first angle, and extract the first angle facial features of the target user from the first facial information.

[0040] The similarity between the first-angle facial features and multiple benchmark facial features in the facial feature database is evaluated to obtain the first-angle score of the target user.

[0041] When the score from the first angle is lower than the score threshold, the second facial features of the target user are obtained based on the second facial information collected from the second angle. The second angle is different from the first angle.

[0042] The first angle score is weighted and fused with the second angle score obtained based on the second angle facial features to obtain the comprehensive score of the target user. When the comprehensive score is not lower than the score threshold, the target user is determined to meet the facial recognition conditions, and facial recognition processing is performed on the target user.

[0043] Fourthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, performs the following steps:

[0044] Acquire first facial information of the target user, take the acquisition angle of the first facial information as the first angle, and extract the first angle facial features of the target user from the first facial information.

[0045] The similarity between the first-angle facial features and multiple benchmark facial features in the facial feature database is evaluated to obtain the first-angle score of the target user.

[0046] When the score from the first angle is lower than the score threshold, the second facial features of the target user are obtained based on the second facial information collected from the second angle. The second angle is different from the first angle.

[0047] The first angle score is weighted and fused with the second angle score obtained based on the second angle facial features to obtain the comprehensive score of the target user. When the comprehensive score is not lower than the score threshold, the target user is determined to meet the facial recognition conditions, and facial recognition processing is performed on the target user.

[0048] Fifthly, this application also provides a computer program product. The computer program product includes a computer program that, when executed by a processor, performs the following steps:

[0049] Acquire first facial information of the target user, take the acquisition angle of the first facial information as the first angle, and extract the first angle facial features of the target user from the first facial information.

[0050] The similarity between the first-angle facial features and multiple benchmark facial features in the facial feature database is evaluated to obtain the first-angle score of the target user.

[0051] When the score from the first angle is lower than the score threshold, the second facial features of the target user are obtained based on the second facial information collected from the second angle. The second angle is different from the first angle.

[0052] The first angle score is weighted and fused with the second angle score obtained based on the second angle facial features to obtain the comprehensive score of the target user. When the comprehensive score is not lower than the score threshold, the target user is determined to meet the facial recognition conditions, and facial recognition processing is performed on the target user.

[0053] The aforementioned facial recognition processing method, apparatus, computer equipment, storage medium, and computer program product use the acquisition angle of the first facial information as the first angle. It extracts the first-angle facial features of the target user from the first facial information, then performs a similarity assessment between the first-angle facial features and multiple benchmark facial features in a facial feature database to obtain a first-angle score for the target user. When the first-angle score is lower than a scoring threshold, it acquires the second facial information of the target user based on a second angle different from the first angle, obtaining the second-angle facial features. Then, it performs a weighted fusion of the first-angle score and the second-angle score obtained based on the second-angle facial features to obtain a comprehensive score for the target user. When the comprehensive score is not lower than the scoring threshold, the target user is determined to meet the facial recognition conditions, and facial recognition processing is performed on the target user. Throughout the process, when the facial information from the first angle does not meet the facial recognition conditions, it acquires facial information of the target user from other angles. This avoids facial recognition failure due to pose deviations at a certain angle. By comprehensively considering multiple angles, the impact of pose deviations at a certain angle can be reduced, thereby improving the success rate of facial recognition. Attached Figure Description

[0054] Figure 1 This is an application environment diagram of the facial recognition processing method in one embodiment;

[0055] Figure 2 This is a flowchart illustrating a facial recognition processing method in one embodiment;

[0056] Figure 3 This is a flowchart illustrating the facial recognition processing method in another embodiment;

[0057] Figure 4 This is a structural block diagram of a facial recognition processing device in one embodiment;

[0058] Figure 5 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0059] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0060] The facial recognition processing method provided in this application embodiment can be applied to, for example... Figure 1 In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be integrated onto server 104 or placed on a cloud or other network server. Server 104 can acquire first facial information collected from the target user through terminal 102, using the acquisition angle of the first facial information as the first angle, and extracting the target user's first-angle facial features from the first facial information. The first-angle facial features are then compared with multiple benchmark facial features in a facial feature library to obtain a first-angle score for the target user. When the first-angle score is lower than a scoring threshold, second facial information is collected from a second angle different from the first angle through terminal 102 to extract the target user's second-angle facial features. The first-angle score and the second-angle score obtained based on the second-angle facial features are weighted and fused to obtain a comprehensive score for the target user. When the comprehensive score is not lower than a scoring threshold, the target user is determined to meet the facial recognition conditions, and facial recognition processing is performed on the target user. Terminal 102 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, and other devices equipped with facial information acquisition functions. Server 104 can be implemented using a standalone server or a server cluster consisting of multiple servers.

[0061] In one embodiment, such as Figure 2As shown, a facial recognition processing method is provided, which is applied to... Figure 1 Taking the server in the example, the following steps are included:

[0062] Step 202: Obtain the first facial information collected from the target user, take the collection angle of the first facial information as the first angle, and extract the first angle facial features of the target user from the first facial information.

[0063] The specific angles for data collection can include the frontal face angle, the left side face angle, and the right side face angle.

[0064] Optionally, after receiving a facial recognition request initiated by the target user through the terminal, the server can acquire the target user's first facial information through the terminal's facial information acquisition device, identify the acquisition angle of the first facial information, use the acquisition angle of the first facial information as the first angle, and then extract the target user's first angle facial features from the first facial information. The acquisition angle of the first facial information can be one of a frontal angle, a left side angle, or a right side angle.

[0065] For example, when the server identifies the first facial information as the frontal face information of the target user, it can use the frontal face angle as the first angle. When the server identifies the first facial information as the left / right side face information of the target user, it can use the left / right side face angle as the first angle.

[0066] Step 204: The first-angle facial features are compared with multiple benchmark facial features in the facial feature library to obtain the first-angle score of the target user.

[0067] The first angle score can be a value between 0 and 100.

[0068] Optionally, the server can perform similarity assessments between the first-angle facial feature and multiple benchmark facial features in the facial feature database to select the benchmark facial feature with the highest similarity to the first-angle facial feature. The score between the benchmark facial feature with the highest similarity and the first-angle facial feature is then used as the target user's first-angle score. The higher the similarity between the first-angle facial feature and the benchmark facial features, the higher the score.

[0069] Step 206: When the score of the first angle is lower than the score threshold, the second angle facial features of the target user are obtained based on the second facial information collected from the second angle. The second angle is different from the first angle.

[0070] The scoring threshold can be configured according to the actual application scenario and is in numerical form.

[0071] Optionally, when the score of the first angle is lower than the score threshold, the server can determine that the facial information of the target user in the first angle does not meet the facial recognition conditions, and select a second angle from the collection angles that are different from the first angle, and then obtain the second facial information of the target user collected from the second angle, so as to obtain the second angle facial features of the target user from the second facial information.

[0072] For example, taking the first angle as a frontal face angle as an example, when the first facial information does not meet the facial recognition conditions, the server can select any one of the following acquisition angles as the second angle: the right face angle or the left face angle.

[0073] For example, taking the first angle as the right face angle / left face angle as an example, the server can use the frontal face angle as the second angle.

[0074] Optionally, when the first angle score is not lower than the score threshold, the server can determine that the first facial information of the target user meets the facial recognition conditions, and perform facial recognition processing on the target user to determine the identity information of the target user.

[0075] Step 208: The first angle score and the second angle score obtained based on the second angle facial features are weighted and fused to obtain the comprehensive score of the target user. When the comprehensive score is not lower than the score threshold, the target user is determined to meet the facial recognition conditions, and facial recognition processing is performed on the target user.

[0076] Each acquisition angle has its own corresponding weight, which can be configured according to the actual application scenario. Taking the acquisition angles including the frontal face angle, left side face angle, and right side face angle as an example, in this embodiment, since the feature points acquired from the frontal face angle are generally more than those acquired from the left / right side face angles, the weight of the frontal face angle is greater than the weight of the left / right side face angle.

[0077] Optionally, if the score from the first angle falls below a scoring threshold, the server can select a second angle. For the identifiable user corresponding to the score from the first angle, the server can obtain the baseline facial features of that identifiable user from the second angle and compare the similarity between these baseline facial features and the target user's facial features from the second angle. The resulting score is then used as the second angle score. Next, the server can weight and fuse the first and second angle scores according to their respective weights to obtain a comprehensive score for the target user. This comprehensive score is then compared with a pre-configured scoring threshold. If the comprehensive score is not lower than the threshold, the target user is deemed to meet the facial recognition criteria, and facial recognition processing is performed to determine the target user's identity information.

[0078] For example, taking a first angle as the frontal view and a second angle as the left side view, and considering a baseline facial feature of identifiable user X at the frontal view that has the highest similarity to the target user's facial feature at the first angle, the server can obtain a score between the baseline facial feature of identifiable user X at the frontal view and the facial feature at the first angle through similarity evaluation, and use the obtained score as the first angle score. When the first angle score is lower than a scoring threshold, and the second angle is the left side view, the server can obtain the baseline facial feature of identifiable user X at the left side view, and perform a similarity evaluation between the baseline facial feature of identifiable user X at the left side view and the target user's facial feature at the second angle, using the obtained score as the second angle score. The first angle score and the second angle score are then weighted and fused to obtain a comprehensive score.

[0079] In the aforementioned facial recognition processing method, the acquisition angle of the first facial information is taken as the first angle. The first-angle facial features of the target user are extracted from the first facial information. Then, the first-angle facial features are compared with multiple benchmark facial features in the facial feature database to obtain a first-angle score for the target user. When the first-angle score is lower than a scoring threshold, second facial information of the target user is obtained based on a second angle different from the first angle, and the second-angle facial features of the target user are obtained. Then, the first-angle score and the second-angle score obtained based on the second-angle facial features are weighted and fused to obtain a comprehensive score for the target user. When the comprehensive score is not lower than the scoring threshold, the target user is determined to meet the facial recognition conditions, and facial recognition processing is performed on the target user. Throughout the process, when the facial information from the first angle does not meet the facial recognition conditions, facial information of the target user from other angles is acquired. This avoids facial recognition failure due to pose deviations at a certain angle. By comprehensively considering multiple angles, the impact of pose deviations at a certain angle can be reduced, thereby improving the success rate of facial recognition.

[0080] In one embodiment, prior to acquiring the first facial information collected from the target user, the process includes:

[0081] Identify multiple identifiable users who meet the facial recognition criteria;

[0082] For each identifiable user, acquire the facial features of the identifiable user from multiple acquisition angles;

[0083] For each acquisition angle, the facial features corresponding to the acquisition angle are used as the reference facial features for the acquisition angle, thus obtaining the reference facial features of the user at each of the multiple acquisition angles.

[0084] A facial feature library is constructed based on multiple baseline facial features for each identifiable user.

[0085] Each baseline facial feature has its own corresponding acquisition angle, and the facial feature library includes multiple facial feature libraries that can identify each user.

[0086] Optionally, the server can pre-determine the identity information of multiple identifiable users who meet the facial recognition conditions. After obtaining authorization from each identifiable user, the server can collect facial information of each identifiable user from multiple acquisition angles to obtain the facial features corresponding to each acquisition angle. Furthermore, for each acquisition angle, the server can use the facial features corresponding to that acquisition angle as the reference facial features for that acquisition angle, obtaining the reference facial features of each identifiable user from multiple acquisition angles. Then, based on the multiple reference facial features of each identifiable user, the server constructs a facial feature library for each identifiable user to obtain the final facial feature library.

[0087] In this embodiment, a facial feature library is constructed using multiple baseline facial features of each identifiable user, so as to perform facial recognition on the target user based on the baseline facial features stored in the facial feature library.

[0088] In one embodiment, each extracted facial feature includes multiple feature points; the first angle facial feature is compared with multiple benchmark facial features in the facial feature library to obtain the target user's first angle score, which includes:

[0089] All benchmark facial features in the facial feature database that were sampled at the same angle as the first angle were used as benchmark features to be compared.

[0090] For each benchmark feature to be compared, multiple feature points in the first angle facial feature are compared with multiple feature points in the benchmark feature to be compared, and the feature point similarity score between the first angle facial feature and the benchmark feature to be compared is obtained.

[0091] The highest feature point similarity score among multiple feature point similarity scores is used as the target user's first angle score.

[0092] Among them, the benchmark features to be compared include: multiple benchmark facial features that can identify each user at the same acquisition angle as the first angle, that is, the benchmark features to be compared include multiple benchmark facial features.

[0093] Optionally, for a facial feature library storing multiple baseline facial features that can identify each user, the server can use all baseline facial features in the facial feature library that were captured at the same angle as the first angle as the baseline features to be compared. For each baseline feature to be compared, the server can evaluate the similarity between multiple feature points in the first angle facial feature and multiple feature points in the baseline feature to be compared. The higher the similarity between the feature points in the first angle facial feature and the feature points in the baseline feature to be compared, the higher the feature point similarity score between the first angle facial feature and the baseline feature to be compared. After obtaining the feature point similarity score between the first angle facial feature and the baseline feature to be compared, the server can use the highest feature point similarity score among the multiple feature point similarity scores as the first angle score for the target user.

[0094] For example, taking a frontal view as the first angle, when there exists a baseline facial feature of identifiable user X at the frontal view, and the similarity score between this baseline feature and the target user's facial feature at the first angle is the highest, the server can initially determine that the target user's identity information is most likely to match the identity information of identifiable user X. The target user is most likely to be identified as the identifiable user; that is, the target user and the identifiable user are most likely to meet the facial recognition conditions. Then, the server can use the feature point similarity score between the target user and the identifiable user X as the target user's first angle score.

[0095] Optionally, when the target user's first angle score is not lower than the score threshold, the server can determine that the target user meets the facial recognition conditions, and the identity information of the identifiable user is consistent with the identity information of the target user. Furthermore, the server can authenticate the target user's identity based on the identity information of the identifiable user.

[0096] Optionally, taking the highest similarity score between the identifiable user X's baseline facial features at a first angle and the target user's facial features at the first angle as an example, when the target user's first angle score is lower than a scoring threshold, the server can initially determine that the identifiable user X and the target user do not meet the facial recognition conditions at the first angle. To comprehensively consider multiple angles, the server can obtain the target user's second facial information based on a second angle different from the first angle, and obtain the target user's second facial features. Then, the similarity between the identifiable user X's baseline facial features at the second angle and the target user's second facial features is compared to obtain the target user's second angle score. Finally, the first angle score and the second angle score are weighted and fused to obtain the target user's comprehensive score, which is used to determine whether the target user and the identifiable user X meet the facial recognition conditions.

[0097] In this embodiment, by obtaining the first angle score of the target user and comparing the first angle score with the score threshold, it is possible to determine whether the identifiable user corresponding to the first angle score and the target user meet the facial recognition conditions under the same acquisition angle as the first angle.

[0098] In one embodiment, when the overall score is not lower than a score threshold, the target user is determined to meet the facial recognition conditions. After performing facial recognition processing on the target user, the process includes:

[0099] The first-angle facial features and the second-angle facial features of the target user are respectively used as the new facial features of the target user;

[0100] For each new facial feature, in the target user's facial feature library, the reference facial feature with the same acquisition angle corresponding to the new facial feature is used as the reference facial feature to be supplemented for the new facial feature.

[0101] From the multiple feature points of the newly added facial features, select the difference feature points that differ from the feature points of the corresponding baseline facial features to be supplemented;

[0102] Based on the acquisition angle corresponding to the newly added facial features, the difference feature points corresponding to the newly added facial features are used as auxiliary facial features for the corresponding acquisition angle, and the auxiliary facial features are added to the target user's facial feature library.

[0103] Optionally, the server can use the first-angle facial features and second-angle facial features of the target user as new facial features. For each new facial feature, in the target user's facial feature library, a baseline facial feature with the same acquisition angle corresponding to the new facial feature is selected as a baseline facial feature to be supplemented. Multiple feature points in the new facial feature are compared with multiple feature points in the corresponding baseline facial feature to be supplemented, thereby selecting difference feature points from the multiple feature points of the new facial feature that differ from the feature points of the corresponding baseline facial feature. Then, based on the acquisition angle corresponding to the new facial feature, the difference feature points corresponding to the new facial feature are used as auxiliary facial features for the corresponding acquisition angle, and these auxiliary facial features are added to the target user's facial feature library to supplement the target user's facial feature library.

[0104] For example, taking the scenario where identifiable user X and target user meet the facial recognition criteria, the facial feature library of identifiable user X is the same as the facial feature library of target user X. The target user's facial feature library stores the baseline facial features corresponding to the target user at each of the frontal, left, and right face angles. Taking the first angle as a frontal angle as an example, the server can use the target user's first-angle facial features as new facial features for the target user at the frontal angle, and use the baseline facial features for the frontal angle in the target user's facial feature library as supplementary baseline facial features for the frontal angle. Then, for the frontal angle, the server can compare multiple feature points from the new facial features with multiple feature points from the corresponding supplementary baseline facial features to obtain multiple difference feature points, thereby obtaining the auxiliary facial features corresponding to the frontal angle. Based on a similar method, the server can obtain the auxiliary facial features corresponding to the first and second angles respectively.

[0105] Among them, the auxiliary facial features only record the feature points in the newly added facial features that differ from the corresponding baseline facial features to be supplemented, and do not record feature points that do not differ, thereby reducing the amount of data in the facial feature library.

[0106] Optionally, when the score of the first angle is not lower than the score threshold, the server does not need to process the second angle. Therefore, when the score of the first angle is not lower than the score threshold, only the auxiliary facial features of the first angle are added to the facial feature library of the target user.

[0107] In this embodiment, after the target user meets the facial recognition conditions, by adding auxiliary facial features of the target user from the first angle and the second angle to the target user's facial feature database, the purpose of supplementing the target user's facial feature database can be achieved.

[0108] In one embodiment, the target user's facial feature library stores auxiliary facial features and baseline facial features of the target user; each auxiliary facial feature includes at least one differential feature point; adding the auxiliary facial features to the target user's facial feature library includes:

[0109] For each acquisition angle, the total amount of auxiliary facial features corresponding to the acquisition angle of the target user is obtained from the target user's facial feature database.

[0110] The acquisition angles where the total data volume exceeds the data volume threshold are taken as acquisition angles to be adjusted, and the difference feature points in multiple auxiliary facial features corresponding to the acquisition angles to be adjusted are compared.

[0111] For each auxiliary facial feature corresponding to the acquisition angle to be adjusted, the comparison result is determined as the total number of different feature points;

[0112] Auxiliary facial features that do not meet the total number requirement are deleted, and the baseline facial features of the target user at the acquisition angle to be adjusted are updated based on the auxiliary facial features with the fewest total number.

[0113] The data volume threshold can be configured according to the actual application scenario. The condition of not meeting the number condition indicates that the total number of auxiliary facial features exceeds the total number threshold, which can also be configured according to the actual application scenario.

[0114] Optionally, after adding the auxiliary facial features of the target user to the target user's facial feature database to supplement the database, for each acquisition angle, the server can obtain the total amount of auxiliary facial features corresponding to the target user at that acquisition angle from the database. Acquisition angles with a total data volume exceeding a threshold are designated as acquisition angles to be adjusted. Then, the server compares the difference feature points among the multiple auxiliary facial features corresponding to the acquisition angle to be adjusted. For each auxiliary facial feature corresponding to the acquisition angle to be adjusted, the server selects the total number of different difference feature points from other auxiliary facial features. This number is determined by the comparison result of each auxiliary facial feature. Finally, auxiliary facial features whose total number does not meet the requirement are deleted, and the baseline facial features of the target user at the acquisition angle to be adjusted are updated based on the auxiliary facial features with the fewest total number of differences.

[0115] For example, after satisfying the facial recognition conditions four times, the total amount of auxiliary facial features corresponding to the frontal face angle in the target user's facial feature database exceeds the data volume threshold (the frontal face angle is the angle to be adjusted). Assume that after the first successful recognition, difference feature points A1, A2, and A3 are recorded in auxiliary facial feature A added at the frontal face angle; after the second successful recognition, difference feature points B1 and B2 are recorded in auxiliary facial feature B added at the frontal face angle; after the third successful recognition, difference feature point C1 is recorded in auxiliary facial feature C added at the frontal face angle; and after the fourth successful recognition, difference feature points D1 and D2 are recorded in auxiliary facial feature D added at the frontal face angle. For auxiliary facial feature A, the server can compare the difference feature points and select those that are different from A1, A2, and A3 from B1, B2, C1, D1, and D2, obtaining the total number 'a' corresponding to auxiliary facial feature A. For auxiliary facial feature B, the server can select differential feature points from A1, A2, A3, C1, D1, and D2 that are different from both B1 and B2, thus obtaining the total number b corresponding to auxiliary facial feature B. Using a similar method, the server can obtain the total number c corresponding to auxiliary facial feature C and the total number d corresponding to auxiliary facial feature D.

[0116] Furthermore, taking a total number threshold of 0, and a > 0 > b > d as an example, the server can determine that the auxiliary facial feature A corresponding to the first recognition does not meet the number condition, meaning that the difference feature points recorded in auxiliary facial feature A are not accurate enough and have significant differences from the difference feature points recorded in other auxiliary facial features. Therefore, the server can delete auxiliary facial feature A to reduce the amount of auxiliary facial feature data stored in the target user's facial feature library at the frontal angle. Then, based on the auxiliary facial feature D corresponding to the fourth recognition, the server can update the target user's baseline facial features at the frontal angle, obtaining the updated baseline facial features of the target user at the frontal angle.

[0117] In this embodiment, considering that the facial features of the target user may change due to factors such as age, by setting a data volume threshold, the baseline facial features corresponding to each collection angle in the facial feature database of the target user can be updated, thereby improving the accuracy and success rate of facial recognition.

[0118] In one embodiment, when the overall score is not lower than a score threshold, the target user is determined to meet the facial recognition conditions. After performing facial recognition processing on the target user, the method further includes:

[0119] Dynamic behavior detection of target users is performed within a preset effective time period;

[0120] When a target user meets the dynamic behavior detection conditions within the valid time, the target user is authenticated based on the identity information obtained by facial recognition processing.

[0121] The effective time can be configured according to the actual application scenario, specifically ranging from 60 seconds to 150 seconds. Dynamic behavior detection includes, but is not limited to, liveness detection and voiceprint detection.

[0122] Optionally, after obtaining the target user's identity information through facial recognition processing, the server can randomly generate dynamic behavior instructions and, within a preset valid time, perform dynamic behavior detection on the target user based on the randomly generated dynamic behavior instructions. When the target user's dynamic behavior within the valid time is consistent with the dynamic behavior instructions, it is determined that the target user meets the dynamic behavior detection conditions, and the target user's identity is authenticated based on the identity information obtained from facial recognition processing.

[0123] For example, dynamic behavioral instructions in liveness detection include, but are not limited to, "nodding," "blinking," and "shaking the head," and these "nodding," "blinking," and "shaking the head" can be randomly combined. Dynamic behavioral instructions in voiceprint detection include, but are not limited to, "reciting multiple randomly generated numbers," to prevent other personnel from using pre-recorded audio.

[0124] In this embodiment, dynamic behavior detection ensures that the facial recognition is performed on the target user, preventing other personnel from using multi-angle photos or pre-recorded videos of the target user for facial recognition, thus further improving the security of identity authentication.

[0125] In another embodiment, such as Figure 3 The diagram shows a flowchart of another facial recognition processing method, which mainly includes the following steps:

[0126] Step 302: Identify multiple identifiable users who meet the facial recognition conditions, and for each identifiable user, acquire the facial features corresponding to each user at multiple acquisition angles.

[0127] Step 304: For each acquisition angle, the facial features corresponding to the acquisition angle are used as the reference facial features of the acquisition angle to obtain the reference facial features of the identifiable user at multiple acquisition angles. Based on the multiple reference facial features of each identifiable user, a facial feature library is constructed.

[0128] Step 306: Obtain the first facial information collected from the target user, take the collection angle of the first facial information as the first angle, extract the first angle facial features of the target user from the first facial information, and take the benchmark facial features in the facial feature library that have the same collection angle as the first angle as the benchmark features to be compared.

[0129] Step 308: For each benchmark feature to be compared, perform a similarity assessment on multiple feature points in the first angle facial feature and multiple feature points in the benchmark feature to be compared, and obtain a feature point similarity score between the first angle facial feature and the benchmark feature to be compared.

[0130] Step 310: Among multiple feature point similarity scores, the highest feature point similarity score is taken as the first angle score of the target user. When the first angle score is lower than the score threshold, the second angle facial features of the target user are obtained based on the second facial information collected from the second angle. The second angle is different from the first angle.

[0131] Step 312: The first angle score and the second angle score obtained based on the second angle facial features are weighted and fused to obtain the comprehensive score of the target user. When the comprehensive score is not lower than the score threshold, the target user is determined to meet the facial recognition conditions and facial recognition processing is performed on the target user.

[0132] Step 314: The first angle facial features and the second angle facial features of the target user are respectively used as the new facial features of the target user. For each new facial feature, the reference facial features with the same acquisition angle corresponding to the new facial feature in the target user's facial feature library are used as the reference facial features to be supplemented for the new facial feature.

[0133] Step 316: From the multiple feature points of the newly added facial features, select the difference feature points that differ from the feature points of the corresponding baseline facial features to be supplemented, and according to the acquisition angle corresponding to the newly added facial features, use the difference feature points corresponding to the newly added facial features as auxiliary facial features for the corresponding acquisition angle, and add the auxiliary facial features to the facial feature library of the target user.

[0134] Step 318: For each acquisition angle, obtain the total amount of auxiliary facial features corresponding to the acquisition angle from the facial feature library of the target user. Acquisition angles with a total amount of data exceeding the data amount threshold are taken as acquisition angles to be adjusted, and the difference feature points among multiple auxiliary facial features corresponding to the acquisition angles to be adjusted are compared.

[0135] Step 320: For each auxiliary facial feature corresponding to the acquisition angle to be adjusted, determine the total number of different feature points in the comparison result, delete the auxiliary facial features that do not meet the number condition, and update the baseline facial features of the target user at the acquisition angle to be adjusted based on the auxiliary facial features with the fewest total number.

[0136] Step 322: Within a preset effective time, perform dynamic behavior detection on the target user. When the target user meets the dynamic behavior detection conditions within the effective time, perform identity authentication on the target user based on the identity information obtained by facial recognition processing.

[0137] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0138] Based on the same inventive concept, this application also provides a facial recognition processing device for implementing the facial recognition processing method described above. The solution provided by this device is similar to the implementation described in the above method; therefore, the specific limitations in one or more facial recognition processing device embodiments provided below can be found in the limitations of the facial recognition processing method described above, and will not be repeated here.

[0139] In one embodiment, such as Figure 4 As shown, a facial recognition processing device is provided, including: a facial feature extraction module 402, a score acquisition module 404, a second angle acquisition module 406, and a comprehensive score acquisition module 408, wherein:

[0140] The facial feature extraction module 402 is used to acquire first facial information collected from the target user, take the acquisition angle of the first facial information as the first angle, and extract the first angle facial features of the target user from the first facial information.

[0141] The scoring module 404 is used to evaluate the similarity between the first-angle facial features and multiple benchmark facial features in the facial feature library to obtain the first-angle score of the target user.

[0142] The second angle acquisition module 406 is used to obtain the second angle facial features of the target user based on the second facial information collected from the target user from the second angle when the first angle score is lower than the score threshold. The second angle is different from the first angle.

[0143] The comprehensive score acquisition module 408 is used to perform weighted fusion of the first angle score and the second angle score obtained based on the second angle facial features to obtain the comprehensive score of the target user. When the comprehensive score is not lower than the score threshold, it is determined that the target user meets the facial recognition conditions, and facial recognition processing is performed on the target user.

[0144] In one embodiment, the facial recognition processing device further includes a facial feature library construction module. The facial feature library construction module is used to determine multiple identifiable users that meet the facial recognition conditions. For each identifiable user, it acquires the facial features corresponding to the identifiable user at multiple acquisition angles. For each acquisition angle, it uses the facial features corresponding to the acquisition angle as the reference facial features of the acquisition angle to obtain the reference facial features of the identifiable user at multiple acquisition angles. Then, based on the multiple reference facial features of each identifiable user, it constructs a facial feature library.

[0145] In one embodiment, each extracted facial feature includes multiple feature points. The scoring module is further configured to use all benchmark facial features in the facial feature library that have the same acquisition angle as the first angle as the benchmark features to be compared. For each benchmark feature to be compared, the multiple feature points in the first angle facial feature are compared with the multiple feature points in the benchmark feature to be compared to obtain a feature point similarity score between the first angle facial feature and the benchmark feature to be compared. Then, the highest feature point similarity score among the multiple feature point similarity scores is taken as the first angle score of the target user.

[0146] In one embodiment, the facial recognition processing device further includes an auxiliary facial feature addition module. The auxiliary facial feature addition module is used to take the first angle facial feature and the second angle facial feature of the target user as new facial features of the target user. For each new facial feature, in the target user's facial feature library, the reference facial feature with the same acquisition angle as the new facial feature is taken as the reference facial feature to be supplemented for the new facial feature. Then, from the multiple feature points of the new facial feature, the difference feature points that are different from the feature points of the corresponding reference facial feature to be supplemented are selected. Finally, according to the acquisition angle corresponding to the new facial feature, the difference feature points corresponding to the new facial feature are taken as auxiliary facial features of the corresponding acquisition angle, and the auxiliary facial features are added to the target user's facial feature library.

[0147] In one embodiment, the target user's facial feature library stores auxiliary facial features and baseline facial features of the target user, with each auxiliary facial feature including at least one differential feature point. The facial recognition processing device also includes a baseline facial feature update module. This module is used to, for each acquisition angle, obtain the total data volume of auxiliary facial features corresponding to the target user at that acquisition angle from the target user's facial feature library, designate acquisition angles with a total data volume exceeding a data volume threshold as acquisition angles to be adjusted, compare the differential feature points among the multiple auxiliary facial features corresponding to the acquisition angles to be adjusted, determine the total number of different differential feature points for each auxiliary facial feature corresponding to the acquisition angles to be adjusted, and finally delete auxiliary facial features whose total number does not meet the requirement. Based on the auxiliary facial feature with the fewest total differential feature points, the baseline facial features of the target user at the acquisition angles to be adjusted are updated.

[0148] In one embodiment, the facial recognition processing device further includes an identity authentication module, which is used to perform dynamic behavior detection on the target user within a preset effective time. When the target user meets the dynamic behavior detection conditions within the effective time, the target user is authenticated based on the identity information obtained by facial recognition processing.

[0149] Each module in the aforementioned facial recognition processing device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module.

[0150] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 5As shown, this computer device includes a processor, memory, input / output interfaces (I / O), and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores facial recognition processing data. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When the computer program is executed by the processor, it implements a facial recognition processing method.

[0151] Those skilled in the art will understand that Figure 5 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0152] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0153] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0154] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0155] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data shall comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0156] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0157] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0158] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A facial recognition processing method, characterized in that, The method includes: Acquire first facial information of the target user, take the acquisition angle of the first facial information as the first angle, and extract the first angle facial features of the target user from the first facial information; The first angle facial features are compared with multiple benchmark facial features in the facial feature library to obtain the first angle score of the target user. When the score from the first angle is lower than the score threshold, the second facial features of the target user are obtained based on the second facial information collected from the second angle, where the second angle is different from the first angle. The first angle score and the second angle score obtained based on the second angle facial features are weighted and fused to obtain the comprehensive score of the target user. When the comprehensive score is not lower than the score threshold, the target user is determined to meet the facial recognition conditions, and facial recognition processing is performed on the target user. The method further includes, after determining that the target user meets the facial recognition conditions, taking the first angle facial features and the second angle facial features of the target user as new facial features; for each new facial feature, taking a reference facial feature with the same acquisition angle as the target user's facial feature library as a reference facial feature to be supplemented; selecting difference feature points from the multiple feature points of the new facial feature that differ from the feature points of the corresponding reference facial feature to be supplemented; and adding the difference feature points as auxiliary facial features of the corresponding angle to the target user's facial feature library according to the acquisition angle corresponding to the new facial feature; wherein the facial feature library stores the target user's reference facial features and multiple... Historically accumulated auxiliary facial features, each of which contains at least one differential feature point; after each addition of a new auxiliary facial feature, for each acquisition angle, if the total amount of auxiliary facial feature data corresponding to that angle exceeds a preset data volume threshold, then that angle is designated as the acquisition angle to be adjusted, and the differential feature points among all auxiliary facial features under that angle are cross-compared; for each auxiliary facial feature corresponding to the acquisition angle to be adjusted, the total number of its differential feature points that differ from those of other auxiliary facial features is determined; if the total number does not meet the preset number condition, then the auxiliary facial feature is deleted, and the baseline facial feature of the acquisition angle to be adjusted is updated based on the remaining auxiliary facial features with the smallest total number.

2. The method according to claim 1, characterized in that, Before acquiring the first facial information collected from the target user, the process includes: Identify multiple identifiable users who meet the facial recognition conditions; For each identifiable user, obtain the facial features of the identifiable user at each of the multiple acquisition angles; For each of the aforementioned acquisition angles, the facial features corresponding to the acquisition angle are used as the reference facial features for that acquisition angle, thereby obtaining the reference facial features of the identifiable user at each of the multiple acquisition angles. A facial feature library is constructed based on multiple baseline facial features of each identifiable user.

3. The method according to claim 1, characterized in that, Each extracted facial feature includes multiple feature points; The step of evaluating the similarity between the first-angle facial features and multiple benchmark facial features in the facial feature database to obtain the first-angle score of the target user includes: All benchmark facial features in the facial feature library that are collected at the same angle as the first angle are used as benchmark features to be compared. For each of the aforementioned benchmark features to be compared, multiple feature points in the first angle facial feature are compared with multiple feature points in the benchmark feature to be compared to obtain a feature point similarity score between the first angle facial feature and the benchmark feature to be compared. The highest feature point similarity score among the multiple feature point similarity scores is taken as the first angle score of the target user.

4. The method according to claim 1, characterized in that, When the comprehensive score is not lower than the score threshold, the step of determining that the target user meets the facial recognition conditions, and then performing facial recognition processing on the target user, further includes: Dynamic behavior detection is performed on the target user within a preset effective time period; When the target user meets the dynamic behavior detection conditions within the effective time period, the target user is authenticated based on the identity information obtained by facial recognition processing.

5. A facial recognition processing device, characterized in that, The device includes: The facial feature extraction module is used to acquire first facial information collected from a target user, take the acquisition angle of the first facial information as a first angle, and extract the first angle facial features of the target user from the first facial information. The scoring module is used to evaluate the similarity between the first angle facial features and multiple benchmark facial features in the facial feature library to obtain the first angle score of the target user. The second angle acquisition module is used to obtain the second angle facial features of the target user based on the second facial information collected from the target user from the second angle when the first angle score is lower than the score threshold. The second angle is different from the first angle. The comprehensive score acquisition module is used to perform weighted fusion of the first angle score and the second angle score obtained based on the second angle facial features to obtain the comprehensive score of the target user. When the comprehensive score is not lower than the score threshold, it is determined that the target user meets the facial recognition conditions, and facial recognition processing is performed on the target user. The device further includes: An auxiliary facial feature addition module is used to: take the first angle facial feature and the second angle facial feature of the target user as new facial features; for each new facial feature, take the reference facial feature with the same acquisition angle as the target user's facial feature library as the reference facial feature to be supplemented; select the difference feature points that differ from the feature points of the corresponding reference facial feature to be supplemented from the multiple feature points of the new facial feature; according to the acquisition angle corresponding to the new facial feature, take the difference feature points as auxiliary facial features of the corresponding angle and add them to the target user's facial feature library; wherein, the facial feature library stores the target user's reference facial features and multiple historically accumulated auxiliary facial features, and each auxiliary facial feature contains at least one difference feature point; The baseline facial feature update module is used to: after each addition of a new auxiliary facial feature, for each acquisition angle, if the total amount of auxiliary facial feature data corresponding to that angle exceeds a preset data amount threshold, then that angle is taken as the acquisition angle to be adjusted, and cross-compare the difference feature points among all auxiliary facial features under that angle; for each auxiliary facial feature corresponding to the acquisition angle to be adjusted, determine the total number of its difference feature points that are different from the difference feature points among other auxiliary facial features; if the total number does not meet the preset number condition, then delete that auxiliary facial feature, and update the baseline facial feature of the acquisition angle to be adjusted based on the one with the smallest total number among the remaining auxiliary facial features.

6. The apparatus according to claim 5, characterized in that, The device also includes a facial feature library construction module, used for: Identify multiple identifiable users that meet the facial recognition conditions; for each identifiable user, obtain the facial features corresponding to the identifiable user at multiple acquisition angles; for each acquisition angle, use the facial features corresponding to the acquisition angle as the reference facial features of the acquisition angle to obtain the reference facial features of the identifiable user at each of the multiple acquisition angles. A facial feature library is constructed based on multiple baseline facial features of each identifiable user.

7. The apparatus according to claim 5, characterized in that, Each extracted facial feature includes multiple feature points; the scoring module is also used for: In the facial feature library, all benchmark facial features with the same collection angle as the first angle are used as benchmark features to be compared. For each benchmark feature to be compared, multiple feature points in the first angle facial feature are compared with multiple feature points in the benchmark feature to be compared to obtain a feature point similarity score between the first angle facial feature and the benchmark feature to be compared. The highest feature point similarity score among the multiple feature point similarity scores is used as the first angle score of the target user.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 4.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 4.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 4.

Images